[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

VMs: An Experiment in Retranscribing the Voynich Manuscript (long)

I've been experimenting with some retranscribing the Voynich Manuscript.
What I mean here by retranscribing is a process of analyzing some of the
EVA glyphs as ligatures obligatorily associating in one combined form two
adjacent glyphs which are in fact logically independent.  Perhaps it would
be better to call this process repartitioning.  For example, how about
taking EVA r to be i + r-flourish, or EVA o to be e + n-flourish, where
r-flourish and n-flourish are independent in the logic of the underlying
system, but belong to the class of flourishes, glyphs that are required to
be attached to a preceding element of some sort.  In this sense flourishes
are an orthographic equivalent of enclitics, which are spoken elements
that have to be attached accentually to a preceding spoken element, like
English auxiliary "have" (when not emphatically or contrastively
stressed):  "I've got it." vs. "No, I *have* got it."

Orthographic ligatures of this nature are not pervasive in modern European
typesetting, but anyone who has been involved in desktop publishing has
probably discovered the special forms for sequences like ff, fl, ffl, fi,
ffi, etc., and students of Latin are aware of ae and oe.  Medievalists
encounter larger sets of ligatures shading into abbreviations in
manuscripts, and students of Sanskrit have to learn an enormous number of
complex Devanagari ligatures.  Finally, modern Roman cursive hands carry
the process to a regularized ligation to an extreme that we don't even
think of in those terms, except perhaps when trying to explain to our
children how to connect following letters to something a b or a v.

I decided to see what happened if I dissolved some of the supposed
ligatures in the VMs.  I began by separating EVA j l m n r into i +
j-flourish, etc.  This idea has a long history in the literature.  I
decided in the interests of simplicity to write these flourishes as just j
l m n r, so below when I refer to r, etc., I mean the r-flourish, etc.,
not EVA r, etc., which I take to represent combinations of EVA i with this
retranscription r.

I also took the rare z letter to be i + m.

Next, I took advantage of a similar train of logic in the observations of
D'Imperio (1978:24), Stolfi (in his comments on stroke harmony), Groves
and others which resolves certain other letters into ligatures of EVA e +
the j l m n r flourishes above. Specifically, I dissolved d into ej, y
into el, g into em, b and o into en, and s into er.  I experimented with
leaving o as o, but this didn't seem to work as well as o > e + n.

I naturally treated c as e, too.  So, ch becomes eh, and sh becomes erh.

Finally, I decided that the gallows letters EVA t k p f looked like two
varieties of tall strokes plus j or m.  I called the two tall strokes t
and k after the EVA transcriptions beginning with them.  I decided to map
EVA t > t + j, EVA k > k + j, EVA p > t + m, and EVA f > k + m.  Again
note that I have two letters, t and k, in my retranscription whose use is
related to the same letters in EVA, but different.

I ultimately decided to leave a as a, and not to dissolve it into e + i.

I left all other letters, i e q x v, unchanged.

There are at least these possibilities of error here:

I may be totally off track in dissolving some or all of these forms into
multiple glyphs.

I may have mis-paired the flourishes as they attach to i and e.   It's
particularly hard to be sure for the j and m flourishes.  It's not clear
whether the relevant distinctions are length or crossing the center of
gravity of the preceding base character.  Also, the b does seem rather
different from o, though the flourish in the b is perhaps simply a more
exuberant version and the same canonical shape.  A similar observation
applies in regard to m and z.

The general difficulties in associating flourishes across different base
letters holds also in regard to dissolving the gallows characters, with
the additional factor that I am on my own here, as I think this has not
been considered before, though I could be wrong.

It may be of interest to see what a sample of retranscribed text looks
like, so here's a standard example:

\li f1r.P1.1;H
\heva f achy s  y k al  ar  at aiin  s ho l  s ho r y  ct hr es  y  k o r
\hrpt kmaeheler elkjail air atjaiiin erhenil erhenirel etjhireer el kjenir
      s ho l d y

I've lined things up to show what produced what.

I apologize for the elkjail air - a chance of the retranscription, not a
profered decipherment!

I have started from the Takeshi Takahashi "H" full transcription as the
base document in performing the retranscription.  I have used an evolving
series of Tcl scripts to do the text manipulation.  I'm using ActiveState
ActiveTcl 8.4.7 on an elderly Pentium III.

Having retranscribed or repartitioned the text, I have made successive
attempts to parse the resulting words into something like consonants and
vowels.  My present interpretation is that the vowels are represented by
the flourish elements

j l m n r

and by


The bulk of the consonants are represented by several series of the form
a?(i+|e+) (to use regular expression notation), or to put it in English,
an optional a followed by a series of one or more i's or e's.  There are
never more than five i's or e's in sequence (remember the last is attached
to a flourish in the EVA analysis), and never more than three e's after an
a, or four i's.  The set:

i, ii, iii, iiii, iiiii
e, ee, eee, eeee, eeeee
ai, aii, aiii, aiiii
ae, aee, aeee

This approach is influenced of course by the observation, e.g., by Jim
Gillogly, that air looks like a single character, which I honor even as I
slice off the r-flourish to convert it into aii.  It also accounts for my
slightly larger count of ae's.  I make it 195 x ae, with 5 x aee and 3 x
aeee.  These are all more common toward the end of the word, though the
curve is not really significant above ae!  (No formal statistical analysis

The relative numbers of these (composite) glyphs vary considerably, of
course.  I don't know if the range in frequencies (1 to 64K) is
reasonable.  I don't know if enough of them are frequent enough to be
reasonable as consonants, or if they are well enough distributed in words
to make sense.  There are > 100 each of ae, ai, aii, aiii, aiiii, e, ee,
eee, eeee, i, iii, and, looking ahead, of k, t, and q.  They definitely
have differential preferences for word position and following vowel.

Also probably consonants are the initial elements of the gallows, which
occur as a?(t|k), or an optional a followed by t or k:

t k at ak

There are a few unassociated a's, mainly final.  The a's before t and k
may be in the same class, but are divided between initial and penultimate

The q occurs initially, of course, with high frequency, and there are a
few x's and v's, including an ax or two.

Because t and k occur both after C and V elements in this classification,
I suspect they are probably potential second elements in consonant
clusters, for which things like r and l (or s and sh) are reasonable in
typical Western European languages.  It is, of course, interesting that
only the j and m "vowels" should occur after these "consonant" elements.
Interesting and perhaps implausible.

The h glyph seems to pattern as a vowel, in that it mostly follows
consonants.  Hence, of course, the frequency of EVA sh (erh).  In
addition, it can follow the j and m vowels, suggesting that it might be
something like an i or u in natural terms.  Or, if Voynichese is Latin,
I'd strongly suspect it of being e and j and m of being a and o (or the

There are two additional factors to consider.  First is my initial abugida
suggestion.  The various probable consonantal glyphs do occur without
following flourishes, sometimes quite commonly, so the possibility exists
that "no flourish" is actually the notation for one of the vowels, and,
conversely, that "some particular flourish" is the notation for "no
following vowel."  This may be an unnecessary complicaiton, but it is
something to keep in mind.

Second, since the only notation for indicating vowels (leaving aside h) is
attaching a flourish to something, and since flourishes in isolation do
not occur, it is likely that one of the "consonants" is a null that exists
to attach vowel flourishes to in order to represent initial vowels or
vowels that follow other vowels.  I think this is a necessary conclusion
if the flourishes are vowels.

If both considerations apply, then the null consonant effectively
represents the default vowel as an independent element.  For example,
suppose the default vowel is target language a and the null consonant is
Voynich script i, then i without a flourish represents the target
language a initially or in vowel + a sequences.  In that case, if Voynich
(flourish) r represents target language e, then ir is target language e in
the same situations.  If Voynich script e is target language t, then
Voynich e alone represents ta and er is te, and so on.

It might be helpful to reconsider the example of the retranscription above
in light of these observations, reducing the glyphs to cover symbols for
phonological class in canonical form:

\hrpt kmaeheler elkjail air atjaiiin erhenil erhenirel etjhireer el kjenir

Here I've represented h as I to suggest possible diphthongs.  This
"canonicalization" doesn't take into account the possibility of the
default vowel or null consonant notations.  If we arbitrarily call Voynich
i a null consonant and let the default vowel be a and the "no following
vowel" marker be l - not particularly plausible, since il thus becomes a
null sequence - this would yield:

\hrpt kmaeheler elkjail air atjaiiin erhenil erhenirel etjhireer el kjenir
      CVICV  CVC

Say, something like:

ronittu trum mu hlape tuite tuiteut tlaiudu t ruteu tuitetat

The substitutions here, made by hand (and so perhaps done incorrectly!),
are completely arbitrary, and are intended only to show that the results
of all this are not (perhaps) altogether implausible.  The repeated
element rendered arbitrarily as tuite- is from erhen from EVA sho.

This is far from a solution, but it does suggest that by looking at the
script in a different way, even a different way based solidly in existing
speculations, we can come up with an analysis which effectively reverses
the standard assumptions about what is a consonant and what is a vowel,
suggesting, I hope, the possibility of fresh perspectives on how to go
about deciphering the Voynich script.

John E. Koontz
To unsubscribe, send mail to majordomo@xxxxxxxxxxx with a body saying:
unsubscribe vms-list